Metricae
Habits, logged from photos
Your camera roll already knows what you did. Metricae reads it on your iPhone, with no cloud and no account, and turns it into a heatmap of your habits.
Download on the App StoreFree · iPhone · No ads, no account
From one photo to a logged day, in 40 seconds
Built like an instrument
No streak anxiety, no confetti. LED heatmaps, LCD readouts and radial gauges, styled after lab hardware. It shows the data and gets out of the way.
Logs from your photos
- Take a photo in the app, or pick one you already have
- The model suggests the activity, one tap confirms it
- Add the time spent and tags like a muscle group or a spot
- Log by hand anything a camera can't see
Learns what your habits look like
- Every photo you confirm teaches it your gym, your trail, your bar
- Suggestions get sharper the more you log
- The engine view shows how sure it is, and why
Good habits and bad ones
- Minimum goals for what you want more of: gym, 3 days a week
- Maximum goals for what you're cutting: 0 late-night snacks
- Avoidance activities flip the heatmap, so clean days light up
Stats, tags and places
- Streak, last session, weekly average and total over any range
- Time per session, with totals and averages
- Places learned from where your photos were taken
- Tags you use at a place come pre-picked the next time you log there
- Every day keeps its photo, so the history is also a photo diary
Private by design
- Everything runs on your iPhone and works offline
- Photos are never uploaded, there is no server
- No account, no analytics, no ads, no purchases
- The App Store privacy label reads Data Not Collected
HOW IT LEARNS
Metricae runs Apple's MobileCLIP, a small image model that fits on a phone. It turns each photo into a fingerprint of 512 numbers that describe what is in it: a gym, a plate, a wave. The model is 22 MB and never changes.
On top sits one thin layer, a linear probe, the idea from OpenAI's CLIP paper: keep the big model frozen and train a small layer to learn new categories from a few examples. Each habit starts from the photos you confirm for it, and every new confirm retrains the layer on your phone in about half a second. It learns what each habit looks like for you, not for anyone else.